feat: 添加贷后计算

apps/carbon/migrations/0002_fingerprint.py

@@ -0,0 +1,33 @@
+# Generated by Django 4.2.27 on 2026-01-16 08:34
+
+from django.db import migrations, models
+import django.utils.timezone
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('carbon', '0001_initial'),
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='Fingerprint',
+            fields=[
+                ('id', models.CharField(editable=False, help_text='主键ID', max_length=20, primary_key=True, serialize=False, verbose_name='主键ID')),
+                ('create_time', models.DateTimeField(default=django.utils.timezone.now, help_text='创建时间', verbose_name='创建时间')),
+                ('update_time', models.DateTimeField(auto_now=True, help_text='修改时间', verbose_name='修改时间')),
+                ('is_deleted', models.BooleanField(default=False, help_text='删除标记', verbose_name='删除标记')),
+                ('fp_hex', models.CharField(max_length=16, unique=True, verbose_name='simhash(hex)')),
+                ('fp_int', models.BigIntegerField(db_index=True, verbose_name='simhash(int)')),
+                ('seg1', models.IntegerField(db_index=True)),
+                ('seg2', models.IntegerField(db_index=True)),
+                ('seg3', models.IntegerField(db_index=True)),
+                ('seg4', models.IntegerField(db_index=True)),
+                ('score', models.FloatField(verbose_name='得分')),
+            ],
+            options={
+                'indexes': [models.Index(fields=['seg1'], name='carbon_fing_seg1_a18a6c_idx'), models.Index(fields=['seg2'], name='carbon_fing_seg2_5c4cfa_idx'), models.Index(fields=['seg3'], name='carbon_fing_seg3_87273b_idx'), models.Index(fields=['seg4'], name='carbon_fing_seg4_60f65c_idx')],
+            },
+        ),
+    ]

apps/carbon/models.py

@@ -1,5 +1,5 @@
 from django.db import models
-from apps.utils.models import CommonBDModel
+from apps.utils.models import CommonBDModel, BaseModel
 
 # Create your models here.
 class Work(CommonBDModel):
@@ -28,3 +28,27 @@ class Work(CommonBDModel):
     dh_file4 = models.ForeignKey("system.file", on_delete=models.SET_NULL, null=True, blank=True, related_name='work_dh_file4')
     dh_file5 = models.ForeignKey("system.file", on_delete=models.SET_NULL, null=True, blank=True, related_name='work_dh_file5')
     dh_file6 = models.ForeignKey("system.file", on_delete=models.SET_NULL, null=True, blank=True, related_name='work_dh_file6')
+
+
+class Fingerprint(BaseModel):
+    fp_hex = models.CharField(
+        max_length=16, unique=True, verbose_name="simhash(hex)"
+    )
+    fp_int = models.BigIntegerField(
+        db_index=True, verbose_name="simhash(int)"
+    )
+
+    seg1 = models.IntegerField(db_index=True)
+    seg2 = models.IntegerField(db_index=True)
+    seg3 = models.IntegerField(db_index=True)
+    seg4 = models.IntegerField(db_index=True)
+
+    score = models.FloatField(verbose_name="得分")
+
+    class Meta:
+        indexes = [
+            models.Index(fields=["seg1"]),
+            models.Index(fields=["seg2"]),
+            models.Index(fields=["seg3"]),
+            models.Index(fields=["seg4"]),
+        ]

apps/carbon/service.py

@@ -59,6 +59,16 @@ def parse_file(file_path:str):
 def get_fingerprint(text):
     return Simhash(text).value
 
-# --- 汉明距离 ---
+MASK_64 = (1 << 64) - 1
-def hamming_distance(a, b):
+
-    return bin(a ^ b).count("1")
+def hamming_distance(a_u, b_s):
+    return ((a_u ^ (b_s & MASK_64)) & MASK_64).bit_count()
+
+
+def split_simhash(fp_int: int):
+    return (
+        (fp_int >> 48) & 0xffff,
+        (fp_int >> 32) & 0xffff,
+        (fp_int >> 16) & 0xffff,
+        fp_int & 0xffff,
+    )

apps/carbon/views.py

@@ -1,17 +1,18 @@
 from django.shortcuts import render
-from .models import Work
+from .models import Work, Fingerprint
 from .serializers import WorkSerializer, WorkCreateSerializer, WorkDqCalSerializer, WorkDhCalSerializer
 from apps.utils.viewsets import CustomModelViewSet
 from rest_framework.decorators import action
 import os
 from django.conf import settings
 import json
-from apps.carbon.service import parse_file
+from apps.carbon.service import parse_file, get_fingerprint, hamming_distance, split_simhash
 import requests
 from rest_framework.exceptions import ParseError
 import re
 from rest_framework.response import Response
-from django.db import transaction
+from django.db import transaction, IntegrityError
+from django.db.models import Q
 # Create your views here.
 
 LLM_URL = "http://106.0.4.200:9000/v1/chat/completions"
@@ -48,6 +49,11 @@ def ask(input:str, p_name:str, stream=False):
             raise ParseError("模型处理错误超过最大token限制")
         return response.json()["choices"][0]["message"]["content"]
 
+def simhash_to_db(n: int) -> int:
+    return n if n < (1 << 63) else n - (1 << 64)
+
+def simhash_from_db(n: int) -> int:
+    return n if n >= 0 else n + (1 << 64)
 
 class WorkViewSet(CustomModelViewSet):
     queryset = Work.objects.all()
@@ -246,14 +252,78 @@ class WorkViewSet(CustomModelViewSet):
         return Response({"total_score": total_score, "data": data})
 
     @staticmethod
-    def parse_files():
+    def parse_files(work: Work):
-        pass
+        contents = []
+        filenames = []
+        for file in [work.dh_file1, work.dh_file2, work.dh_file3, work.dh_file4, work.dh_file5, work.dh_file6]:
+            if file:
+                if file.name in filenames:
+                    continue
+                path = (settings.BASE_DIR + file.path).replace('\\', '/')
+                content = parse_file(path)
+                filenames.append(file.name)
+                contents.append(content)
+        return '\n'.join(contents)
 
     @action(detail=True, methods=['post'], serializer_class=WorkDhCalSerializer)
     @transaction.atomic
     def cal_dh(self, request, pk):
         work = self.get_object()
-        
+        sr = WorkDqCalSerializer(work, data=request.data)
+        sr.is_valid(raise_exception=True)
+        sr.save()
+        work = Work.objects.get(pk=pk)
+        content = WorkViewSet.parse_files(work)
+
+        fp_u = get_fingerprint(content)          # unsigned
+        fp_int = simhash_to_db(fp_u)              # signed for db
+        fp_hex = format(fp_u, "016x")
+
+        s1, s2, s3, s4 = split_simhash(fp_int)
+
+        # 1️⃣ 分段粗筛
+        candidates = (
+            Fingerprint.objects
+            .filter(
+                Q(seg1=s1) |
+                Q(seg2=s2) |
+                Q(seg3=s3) |
+                Q(seg4=s4)
+            )
+            .only("fp_int", "score")
+        )
+
+        # 2️⃣ 精确海明距离
+        for obj in candidates:
+            if hamming_distance(fp_u, obj.fp_int) <= HAMMING_THRESHOLD:
+                work.score_dh = obj.score
+                work.save(update_fields=["score_dh"])
+                return Response({"total_score": obj.score})
+
+        # 3️⃣ 未命中 → 调用 AI
+        res = ask(content, "tec_dh")
+        score = round(float(res), 2)
+
+        work.score_dh = score
+        work.save(update_fields=["score_dh"])
+
+        # 4️⃣ 并发安全写入指纹库
+        try:
+            Fingerprint.objects.create(
+                fp_hex=fp_hex,
+                fp_int=fp_int,
+                seg1=s1,
+                seg2=s2,
+                seg3=s3,
+                seg4=s4,
+                score=score,
+            )
+        except IntegrityError:
+            # 并发下已存在，忽略即可
+            pass
+
+        return Response({"total_score": score})
+